Exit polls went wild in 2014, were wrong in 2019 too

Pollsters and psephologists in the country can get the leading party right more or less, but all their other numbers generally go way off the mark

author-image
Surajit Dasgupta
New Update
Exit polls how accurate will they be

Exit polls: How accurate will they be in 2024?

Listen to this article
0.75x 1x 1.5x
00:00 / 00:00

On Saturday, 1 June, starting from 5 PM, various exit polls will begin broadcasting live. It's important to note that these are not actual results of the 2024 Lok Sabha elections but rather predictions.

Historically, many exit pollsters have not been very accurate in their predictions. When analysing these polls, it's better to focus on identifying trends rather than specific numbers. 

The Squirrels will share its predictions through articles and videos at 7 PM tonight. While they won't be traditional exit polls, our data will reflect trends that we have observed throughout the campaign. As of now, we recall the past inaccuracies of well-known psephologists.

The accuracy of exit poll predictions for the Lok Sabha elections in India has varied wildly across different years and among different pollsters and psephologists. Here’s an overview of the accuracy — rather inaccuracy — of the exit polls for the 2014 and 2019 Lok Sabha elections:

Exit polls for 2014 Lok Sabha elections

In the 2014 Lok Sabha elections,

1. Times Now-CVoter: Predicted 289 seats for the NDA, with the actual result being 336 seats. The error margin was 47 seats. Percentage Error (PE): 13.99% [Formula: (Absolute error/Actual result) x 100]

2. ABP-Nielsen: Predicted 281 seats for the NDA. The error margin was 55 seats. PE  16.37%

3. India Today-Cicero: Predicted 272-282 seats for the NDA. The error margin ranged from 54 to 64 seats.  PE: 17.56%

4. CNN-IBN-Lokniti CSDS: Predicted 270-282 seats for the NDA. The error margin ranged from 54 to 66 seats. PE: 17.86%

5. News 24-Chanakya: Predicted 340 seats for the NDA, which was the closest with an error margin of 4 seats. PE: 1.19%

Exit polls for 2019 Lok Sabha elections

For the 2019 Lok Sabha elections, exit polls were generally more accurate, with several pollsters predicting results close to the actual outcome. Here are some notable examples:

1.       Times Now-VMR: Predicted 306 seats for the NDA, with the actual result being 353 seats. The error margin was 47 seats. PE: 13.31%

2.       India Today-Axis My India: Predicted 339-365 seats for the NDA. The actual result was 353 seats, placing their prediction within the range. PE: 3.40%

3.       News18-IPSOS: Predicted 336 seats for the NDA, resulting in an error margin of 17 seats. PE: 4.82%

4.       Republic TV-Jan Ki Baat: Predicted 305 seats for the NDA. The error margin was 48 seats. PE: 13.60%

5.       News24-Chanakya: Predicted 350 seats for the NDA, which was very close with an error margin of 3 seats. PE: 0.85%

Prominent psephologists, pollsters

·       Yogendra Yadav (Lokniti-CSDS): Known for his detailed analysis and predictions

·       Pradeep Gupta (Axis My India): His organization was notably accurate in both 2014 and 2019

·       Sanjay Kumar (CSDS): Often featured in media analysis for his insights

·       Madhav Khosla and Suhas Palshikar: Political analysts and psephologists who provide deeper insights into electoral trends

The exit polls for the 2014 Lok Sabha elections had a wider range of inaccuracies, with some pollsters grossly underestimating the BJP-led NDA's performance. In contrast, the 2019 exit polls were more accurate, with many pollsters providing predictions closer to the actual results.

A study of exit polls up to November 2014 showed that pollsters in the country could more or less get the leading party right, but all their other numbers were way off the mark. A few graphs from that study follow.

Graph 1

Graph 2

Here's what a pollster must do to get his calculations closer to the actual election results.

  • Learn from mistakes of the past
  • Do not suffer from confirmation bias
  • Talk more to people, not of your socio-economic class; talk to the poor more because they vote in larger numbers
  • Improve your polling methodology with questions that will get quantifiable answers from those surveyed
  • Increase your sample size
  • Do better data analysis
Exit polls elections